Investıgatıon of solar chimney system performance: experimental and numerical analysis

In this study, the efficiency of the prototype solar chimney built in Batman province was investigated experimentally and numerically in the climatic conditions of the region. Temperature, velocity, radiation and turbine velocity (RPM) were selected as experimental measurement parameters. For detailed analysis of the system, measurements were taken simultaneously in four different directions (north, south, east and west) for 24 hours. Measurements were taken at 24 different points for temperature and 6 different points for velocity at the entrance, middle and exit points of the system. Correlation analyzes were made between the data obtained with the SPSS 17 statistical program. It has been observed that the irradiance values are associated with high correlations with the parameters affecting system performance (temperature, velocity and RPM). It is seen that the correlation value between turbine velocity (RPM) and irradiance values is 0,82 according to 0,01 significant significance level. In this case, it has been determined that there is a high level positive relationship between the parameters. With the regression applied to the experimental parameters, the independent variables that are effective on the dependent variables were determined and the ineffective variables were eliminated in the modeling process. In addition, empirical correlations are derived between the parameters obtained with MATLAB Curve Fitting. The coefficient of coefficient of determination between turbine inlet velocity and ambient temperature and radiation value was found to be 0,97. Nusselt and Rayleigh numbers were calculated to determine the flow type. The effective parameters and power generation values determined as a result of the analysis were used in the training and testing phases of artificial neural network models. The data obtained were trained in a multi-layer ANN model using a feed forward backward propagation Levenbergt Marquert algorithm with three inputs and one output using the MATLAB program. Training and test data with different transfer functions (TANSIG, PURELIN and LOGSIG) with different neuron numbers in 2 layers (5,10,15,20 and 25) with the most appropriate R value were determined. The results obtained after training the ANN model showed that the network was quite successful in predicting the experimental data. In the study, it was observed that the regression value (R) for the LOGSIG transfer function during the training of the network was 0,99. The geometric features of the solar chimney prototype have also been analyzed numerically by using the computational fluid dynamics (CFD). Continuity, momentum and energy equations have been applied to the solar chimney system using the finite volume method. Numerical model was built with the commercial software program ANSYS FLUENT 18.1, then the numerical analysys was performed. The geometry of the prototype has been drawn in three dimensions. Since the flow in the system is turbulent RNG k-epsilon turbulence model is used. In addition, DO (discrete ordinates) model was included in the analysis to evaluate the radiation effect within the collector area. Finally, when the numerical and experimental results were compared, it was seen that the obtained data and numerical data were close to each other